Chapter 2 Paying the price for environmentally sustainable and
2.7 Appendix
2.7.1 A. Product mapping
Table 2.3 Product mapping for groups of food products targeted in the diet scenarios by model.
CAPRI* GLOBIOM* MAGNET#
Vegetables
& fruits Tomatoes Other vegetables Apples
Other fruits Nuts Citrus fruits Grapes Olives
Potatoes Sweet potatoes Pulses
Vegetables+ Fruit Nuts
Edible roots and tubers Pulses
Red &
processed meat
Pork Beef
Sheep and goat meat
Pork Beef
Sheep and goat meat
Cattle
Sheep, goats, horses Pig & other animal products
Processed meat - beef Processed meat - sheep, goats, horse Processed meat - poultry
Processed meat - pork
Sugar Sugar Sugar beet and cane Sugar+
Note: *Products in primary equivalents; #Products in dollar values; +amounts of fruit & vegetables and sugar consumed via processed food are captured when assessing private consumption but not targeted via taxes on the single category of processed food.
2.7.2 B. Supplementary model information
Three established economic models are applied in this study to assess required price changes for reaching nutritionally motivated diet changes on EU level and their implied economic and environmental impacts:
CAPRI (https://www.capri-model.org/)
GLOBIOM (https://www.globiom.org)
MAGNET (https://www.magnet-model.org/).
The business-as-usual (BAU) scenario is based on the “REF0” scenario in Frank et al. (2018). The shared scenario assumptions build on GDP and population developments from SSP2 in the IIASA SSP database (https://tntcat.iiasa.ac.at/SspDb/). Despite that long-term GDP and population projections are aligned and input data such as consumption trends or yield shifts are exchanged between the models, differences remain in the results. These arise from divergencies in model databases and underlying assumptions. The models are calibrated using historical data and future trend projections of various other sources regarding certain parameters. In some cases, problems with the historical data have created surprising kinks in model projections, such as sugar consumption in Czech Republic. In the CAPRI database Czech sugar consumption is strongly declining after 2010 up to 2014 while CAPRI projects a recovery and monotonic increase thereafter. The Czech ex-post data in the database may be questionable but they imply that what looks like a kink in 2030 is in fact a correction of short run fluctuations. If we had taken a seven-year average around 2010 as our starting point instead the CAPRI projections for Czech sugar consumption would have looked monotonic. Furthermore, inflection points can arise due to changes in macro drivers (population and GDP) in the 2010-2030 period versus the 2030-2050 period. Interpolating these points would provide a smoother picture but changes in trends for certain countries would remain.
Based on each models’ database the common drivers in all three models are assumptions of neoclassical economic theory on demand and supply governed by prices.
Demand elasticities in the models are informed by the literature for each of the models as shown in Table 2.4. Average EU expenditure and own-price elasticities for the target products give an impression of direction and magnitude. The elasticities are used in the models in a more detailed
representation varying across EU member states. For MAGNET and CAPRI also cross-price elasticities are included in the modelling systems. In the calibration procedure elasticities are adapted so that constraints implied by economic theory are fulfilled, while keeping the deviation to the exogenous elasticities from the literature as small as possible. Applying elasticities for broad aggregated food products and country averages has the limitation that variation within product and population groups is lost in aggregation. Also, some of the elasticity sources have been published some years previous to the study at hand so that recent consumption trends could not be captured.
As the same holds for most of the data underlying in the models, also the projected per capita consumption trends presented in Figure 2.2 may deviate if more recent information would show a persistent change from past trends.
Furthermore, the applied elasticities are not validated for the imposed consumption shifts in the range analyzed in this study. We found similar tendencies of unenforceable high price shifts necessary to achieve a substantial consumption change on population level across the three economic models. We refrain from conducting a sensitivity analysis on our demand elasticities as this is only one, despite an important, element in modelling systems calibrated to very different settings than the pursued one.
Instead, we want to stress the need for further research on how large-scale diet changes are achievable and in how far non-price interventions could affect demand elasticities and facilitate recommended consumption changes.
Based on this information, future economic modelling projections can be better informed, more flexible and robust.
Table 2.4 EU demand elasticities for groups of food products targeted in the diet scenarios by model
EU demand elasticities
CAPRI GLOBIOM MAGNET
Underlying sources
Muhammad et al. (2011)
Muhammad et al. (2011), Alexandratos and Bruinsma (2012)
Aguiar et al.
(2016)+
Expenditure elasticities
Vegetables & fruits 0.2519 -0.29 -0.001
Red & processed
meat 0.4236 0.05 -0.001
Sugar 0.0004 0.12 -0.001
Own-price elasticities
Vegetables & fruits -0.2509 -0.18 -0.01 (-0.63)*
Red & processed
meat -0.3737 -0.28 -0.65
Sugar -0.15 -0.26 -0.54 (-0.63)*
Note: Calibrated model elasticities are presented as weighted EU averages for the respective product groups +Additional calibration of income elasticities is done in the MAGNET baseline to (1) improve response in demand pattern to strong income increases for current low-income regions by linking the income elasticities to real income per capita; (2) respect physical limitations on calorie consumption for current high-income regions by capping income elasticities at -0.001.
Figures in the table are calibrated MAGNET values (weighted average using base year consumption values across EU28 regions for own price elasticities which vary by EU region).
*Consumption of vegetables & fruits and sugar through processed food are also included in diet target, value in parentheses is the own price elasticity of processed food.
Due to the respective demand system specifications it was not possible for all models to enforce the envisaged diet changes via price interventions. In MAGNET the targeted increase for fruits and vegetables could not be achieved through a subsidy and a taste shifter has been employed instead.
The technical reason is a low price elasticity for vegetables and fruits in the
EU region. These elasticities are estimated using an implicit, directly additive demand system (AIDADS) on GTAP data. While these appear low compared to other sources there is no immediate direction in which to change the elasticities without also considering the flow of vegetables and fruits through other food (which has a high elasticity of up to 0.75). In CAPRI the targeted reduction of sugar intake could not be reached as consumption quantities would have fallen below the minimum consumption levels, which are price- and income- independent elements of the generalized Leontief demand system as specified in the calibration procedure to ensure a diversified consumption bundle. For sugar demand, this parameter varies between 10 and 60 kg/cap/year (before deducting losses) for the EU member states in the underlying CAPRI calibration. This is very close to the simulated consumption, meaning that the largest part of sugar consumption is not responsive to prices in CAPRI, in line with the assumption that total sugar consumption is very inelastic. An additional reason for divergences results from the ‘Armington (1969) approach’ that considers domestic and foreign products to be of different qualities. In the baseline units are chosen such that consumption in quality corrected units is identical to physical units (tons). But in scenarios the quality corrected consumption in CAPRI differs from the physical consumption that is recalculated from the model solution. While the quality corrected consumption is very close to the envisaged target, after conversion into tons the results can deviate from the target. Integrating a parallel physical accounting into the model might be possible, but this will further increase the already long time needed to solve the model. The reported calories from MAGNET are derived from ex-post calculations accounting for two main channels through which primary products reach consumers (direct consumption and processed food). For all aggregated sectors, i.e. sectors where the MAGNET representation covers a wide variety of products, calorie contents per unit of product may vary considerably across countries.
As a result, changes in trade flows may alter calorie content of purchased products, even if the total amount of product remains the same. As these calculations are ex-post we cannot target them in the scenarios, instead relying on a model variable measuring calorie contents without capturing trade-induced changes. While reporting ex-post numbers can result in
discrepancies with targeted amounts, the ex-post calculations provide a more precise measure of calories by respecting the global balance in calories produced and demanded.
Accounted calorie values diverge systematically between the models. Daily calories available per capita on EU average range between 2441 kcal in GLOBIOM and 3776 kcal in CAPRI for BAU 2010 (see supplementary data, Appendix E). While further data improvements (and with it potential improvements in data consistency between models) are needed to validate and improve the representation of the consumption side in these large-scale economic models, the existing structures suffice for a comparable scenario implementation in relative terms.
As the representation of diets and nutrition is coarse in the large-scale economic models, we refer to the 955 FoodEx2 consumer products (including processed products with mixed ingredients) in the SHARP database for these indicators (Mertens et al. 2019). FoodEx2 is the second version of the standardized food classification and description system of the European Food Safety Authority (https://www.efsa.europa.eu/en/data/data-standardisation).
We refer to the nutrition score NRD9.3 to assess the impact on average nutrition quality for three EU member states calculated by MAGNET-SHARP. The improved scores arising from the food pattern (FP) scenario are close to the upper boundary of the range of nutritional differences observed within these populations. We calculate the normalized NRD9.3 based on the data provided by Mertens et al. (2018) and compare it to the average scores in the BAU and FP scenarios (Table 2.5).
Table 2.5 NRD9.3 comparison based on MAGNET-SHARP and observed population range
NRD9.3 Czech Republic Denmark France
BAU 2050 0.6 0.65 0.57
FP 2050 0.71 0.73 0.61
Observed
population range, normalized*
0.46-0.58 0.53-0.67 Not available
Note: *The observed population range was calculated based on the NRD9.3 values for the 25th and the 75th percentile provided in Table 4 in Mertens et al. (2018). To retrieve normalized values between 0 and 1, we used the following formula: 𝑁𝑅𝐷9.3 = (𝑁𝑅𝐷9.3 + 300) 1200⁄ .
Food intake in SHARP excludes food waste and loss shares. There is no explicit tracking of food loss and waste in the economic models. Food loss and waste shares in the economic models are informed by the FAO Food Balance Sheets (http://www.fao.org/faostat/en/#data/FBSH) and partly by FAO (2011). For the MAGNET-SHARP mapping, food loss and waste shares are assumed to be constant over time.
References
Aguiar, A., Narayanan, B., McDougall, R., 2016. An Overview of the GTAP 9 Data Base. J. Glob. Econ. Anal. 1, 181–208.
https://doi.org/10.21642/JGEA.010103AF
Alexandratos, N., Bruinsma, J., 2012. World agriculture towards 2030/2050:
the 2012 revision. FAO. ESA Working paper 12-03. Rome, FAO.
Armington, P.S., 1969. A Theory of Demand for Products Distinguished by Place of Production. Staff Papers - International Monetary Fund 16, 159-178. https://doi.org/10.2307/3866403
FAO, 2011. Global food losses and food waste – Extent, causes and prevention. Rome.
Frank, S., Havlík, P., van Dijk, M., Achterbosch, T., Cui, D., Heckelei, T., Kuiper, M., Latka, C., Witzke, H.-P. (2018): Quantified future challenges to sustainable food and nutrition security in the EU (No.
10.2). SUSFANS project H2020/ SFS-19- 2014: Sustainable food and nutrition security through evidence based EU agro-food policy, GA no.
633692.
Mertens, E., Kuijsten, A., Dofková, M., Mistura, L., D’Addezio, L., Turrini, A., Dubuisson, C., Favret, S., Havard, S., Trolle, E., Van’t Veer, P., Geleijnse, J.M., 2018. Geographic and socioeconomic diversity of food and nutrient intakes: a comparison of four European countries. Eur. J.
Nutr. https://doi.org/10.1007/s00394-018-1673-6
Mertens, E., Kuijsten, A., van Zanten, H.H.E., Kaptijn, G., Dofková, M., Mistura, L., D’Addezio, L., Turrini, A., Dubuisson, C., Havard, S., Trolle, E., Geleijnse, J.M., van ’t Veer, P., 2019. Dietary choices and environmental impact in four European countries. J. Clean. Prod. 237, 117827. https://doi.org/10.1016/j.jclepro.2019.117827
Muhammad, A., Seale, J.L., Meade, B., Regmi, A., 2011. International Evidence on Food Consumption Patterns: An Update Using 2005 International Comparison Program Data. USDA-ERS Technical Bulletin (No. 1929). https://dx.doi.org/10.2139/ssrn.2114337
2.7.3 C. Calorie reduction target calculation
Overweight and obesity are the result of an imbalance between energy intake and energy use for maintenance, growth and physical activity. Working towards a population level policy which is rough by design, we average variations in age, weight, physical activity, and sex. We refer to an EU median age of 43 years in 2018 (Eurostat, 2019) and a mean height of 1.75 m for male and 1.65 m for female adults (Roser et al., 2020). For the four nutritional status groups ‘underweight’, ‘normal’, ‘pre-obese’ and ‘obese’
we make assumptions on the respective average body mass index (BMI), as we are missing information on the distribution of BMI within each group (see Table 2.6). Given the formula for calculating the BMI as dividing a person’s weight in kg by the square of the person’s height in m, we determine the corresponding weight for each nutritional status group for male and female. On this ground, we calculate the basal metabolic rate, the daily total energy expenditure of a person, using the equations provided by FAO (2004) and DAG (2014). To derive the daily calorie requirement for each group we assume a moderately active lifestyle with a physical activity level (PAL) of 1.75 on average and use a conversion from MJ to kcal of 1:239 (DAG, 2014).
The resulting calorie requirements are weighted with the population share of each nutritional status group in the EU in 2017 (Eurostat, 2020). The weighted average calorie requirement exceeds the requirement of the group with a ‘normal’ BMI by about 10%, so that we approximate this divergence as the relative calorie reduction target used for the model assessment.
Table 2.6 Calorie requirement for the EU population average by nutritional status and sex
Nutritional status
Avg.
BMI*
(kg/m²)
Weight (kg) male | female
Energy requ.+ (kcal/
day) male | female
EU pop.
share# (%) male | female
Energy requ. X EU pop.
share male | female underweight 18 57 | 49 2673 |
2177 2 | 5 43 | 109
normal 21 67 | 57 2864 |
2293
40 | 51 1134 | 1162
pre-obese 28 89 | 76 3246 |
2590
43 | 30 1409 | 772
obese 37 117 | 101 3848 |
3042
15 | 15 589 | 444 Weighted EU avg. energy requ. (kcal/day) 3174 |
2487 Excess energy intake on EU avg. relative to energy requ. of
‘normal’ (%)
11 | 8
Note: *As no information on BMI distribution within categories is available, we made assumptions of a potential average (avg.) BMI in these groups. +Energy requirements (requ.) are calculated based on the BMR formula in FAO (2004) for groups with ‘underweight’ and ‘normal’ nutritional status, and on the BMR formula in DAG (2014, p. 79) for groups with ‘pre-obese’ and ‘obese’
nutritional status. We assume a PAL of 1.75 on population (pop.) average. #Based on Eurostat (2020).
References
DAG, 2014. Interdisziplinäre Leitlinie der Qualität S3 zur „Prävention und
Therapie der Adipositas“. Deutsche Adipositas Gesellschaft. URL:
https://www.adipositas-gesellschaft.de/fileadmin/PDF/Leitlinien/S3_Adipositas_Praevention_
Therapie_2014.pdf (accessed 22 May 2020)
Eurostat, 2020. Person distribution by body mass index, educational attainment level, sex and age [ilc_hch10]. URL:
http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=ilc_hch10&la ng=en (accessed 22 May 2020)
Eurostat, 2019. Population structure and aging. URL:
https://ec.europa.eu/eurostat/statistics-explained/index.php/Population_structure_and_ageing#Median_age_i s_highest_in_Italy (accessed 22 May 2020)
FAO, 2004. Human energy requirements Report of a Joint FAO/WHO/UNU Expert Consultation, Rome, 17-24 October 2001. FAO Food and Nutrition Technical Report Series. UNU/ WHO/FAO. URLs:
http://www.fao.org/3/y5686e/y5686e08.htm (accessed 22 May 2020);
http://www.fao.org/3/y5686e/y5686e07.htm#TopOfPage (accessed 22 May 2020)
Roser, M., Appel, C., Ritchie, H., 2020. Human height. Published online at OurWorldInData.org. URL: https://ourworldindata.org/human-height (accessed 22 May 2020)
2.7.4 D. Meat consumption changes
2.7.5 E. Supplementary data description
The description of the supplementary data provided with this article can be found in the online supplementary material of the published article under https://doi.org/10.1016/j.gfs.2020.100437.
Figure 2.6 Percentage consumption changes for red and processed meat in EU member states relative to the business-as-usual scenario in 2010.
Note: The consumption shocks are imposed on EU average level and the models are not constrained to solve for an even distribution across EU member states. The displayed EU member states therefore may show a deviating pattern compared to the EU average change.
Country names according to ISO 3166 Alpha-3 codes.
2.7.6 F. Supplementary data
The supplementary data can be found in the online supplementary material of the published article under https://doi.org/10.1016/j.gfs.2020.100437.